In [6]:
import matplotlib.pyplot as plt
import numpy as np
def imshow(img):
img=img/2+0.5
npimg=img.numpy()
plt.imshow(np.transpose(npimg, (1, 2, 0)))
plt.show()
imshow(torchvision.utils.make_grid(images))
import torch
import torch.nn as nn
import torch.nn.functional as F
import torchvision
#### Model Parameters
dataset=torchvision.datasets.CIFAR100
batch_size = 256
num_epochs = 10
num_workers = 4 # You can use multiple processes to load the data. 0 means that everything is done in the same process
data_dir='CIFAR100_Data' # Folder that contains the dataset. If it is not present, the dataset will be downloaded into that folder
start_epoch=1
device = "cuda" if torch.cuda.is_available() else "cpu"
print(device)
kwargs = {'num_workers': num_workers, 'pin_memory': True} if device=='cuda' else {}
kwargs
cuda
{'num_workers': 4, 'pin_memory': True}
transform = torchvision.transforms.Compose([
# resize
torchvision.transforms.Resize(32),
# center-crop
torchvision.transforms.CenterCrop(32),
# to-tensor
torchvision.transforms.ToTensor(),
# normalize
torchvision.transforms.Normalize([0.5, 0.5, 0.5], [0.5, 0.5, 0.5])
])
train_set=dataset(root=data_dir,train=True,download=True,transform=transform)
test_set=dataset(root=data_dir,train=False,download=True,transform=transform)
Files already downloaded and verified Files already downloaded and verified
train_size=int(0.8 * len(train_set))
val_size=len(train_set)-train_size
train_dataset,validation_dataset=torch.utils.data.random_split(train_set,[train_size, val_size])
train_dataloader=torch.utils.data.DataLoader(train_dataset,batch_size=batch_size,shuffle=True,**kwargs)
validation_dataloader = torch.utils.data.DataLoader(validation_dataset,batch_size=batch_size,shuffle=True,**kwargs)
test_dataloader=torch.utils.data.DataLoader(test_set,batch_size=batch_size,shuffle=False,**kwargs)
print(train_size)
print(val_size)
print(len(test_set))
40000 10000 10000
data_iterator_train=iter(train_dataloader)
images,labels=data_iterator_train.next()
print(images.shape)
print(images.device)
torch.Size([256, 3, 32, 32]) cpu
import matplotlib.pyplot as plt
import numpy as np
def imshow(img):
img=img/2+0.5
npimg=img.numpy()
plt.imshow(np.transpose(npimg, (1, 2, 0)))
plt.show()
imshow(torchvision.utils.make_grid(images))
import torch.optim as optim
from torch.autograd import Variable
import time
import numpy as np
from torch.optim import lr_scheduler
use_cuda = torch.cuda.is_available()
def train_one_epoch(epoch,model, trainloader, optimizer, criterion):
model.train()
train_loss = 0
train_accuracy = 0
for batch_idx, (inputs, targets) in enumerate(trainloader):
inputs, targets = inputs.to(device), targets.to(device)
optimizer.zero_grad()
outputs = model(inputs)
loss = criterion(outputs, targets)
loss.backward()
optimizer.step()
train_loss += loss.item()
accuracy = get_accuracy(outputs, targets)
train_accuracy += accuracy[0].item()
avg_train_loss = train_loss/(batch_idx+1)
avg_train_acc = train_accuracy/(batch_idx+1)
print ('\n=> Training Epoch [%3d/%3d] \tTraining Loss: %.4f%% Training Accuracy: %.3f%%'
%(epoch, num_epochs,avg_train_loss,avg_train_acc))
return avg_train_loss, avg_train_acc, model
def validate(epoch,net, valoader, criterion):
val_loss = 0
val_accuracy = 0
net.eval()
y_pred=[]
with torch.no_grad():
for batch_idx, (inputs, targets) in enumerate(valoader):
inputs, targets = inputs.to(device), targets.to(device)
outputs = net(inputs)
loss = criterion(outputs, targets)
val_loss += loss.item()
accuracy = get_accuracy(outputs, targets)
val_accuracy += accuracy[0].item()
avg_loss = val_loss/(batch_idx+1)
avg_acc = val_accuracy/(batch_idx+1)
print(" | Validation Epoch #%d\t\tValidation Loss: %.4f%% Validation Accuracy: %.2f%%" % (epoch, avg_loss, avg_acc))
return avg_loss, avg_acc
def get_accuracy(output, target, topk=(1,)):
"""Computes the accuracy over the
k top predictions for the specified values of k"""
with torch.no_grad():
maxk = max(topk)
batch_size = target.size(0)
_, pred = output.topk(maxk, 1, True, True)
pred = pred.t()
correct = pred.eq(target.view(1, -1).expand_as(pred))
res = []
for k in topk:
correct_k = correct[:k].view(-1).float().sum(0, keepdim=True)
res.append(correct_k.mul_(100.0 / batch_size))
return res
def start_train_valid(net,trainloader, validloader, criterion):
elapsed_time = 0
lr=0.1
train_loss,train_accuracy,validation_loss,validation_accuracy=[],[],[],[]
for epoch in range(start_epoch,start_epoch+num_epochs):
start_time = time.time()
optimizer=optim.SGD(net.parameters(),lr, momentum=0.9)
trainloss,trainaccuracy,net = train_one_epoch(epoch,net,trainloader,optimizer,criterion)
validationloss,validationaccuracy = validate(epoch,net,validloader,criterion)
train_loss.append(trainloss)
validation_loss.append(validationloss)
train_accuracy.append(trainaccuracy)
validation_accuracy.append(validationaccuracy)
epoch_time = time.time() - start_time
elapsed_time += epoch_time
print(' | Elapsed time : %d:%02d:%02d' % (get_hms(elapsed_time)))
return train_loss, train_accuracy,validation_loss, validation_accuracy
def get_hms(seconds):
m, s = divmod(seconds, 60)
h, m = divmod(m, 60)
return h, m, s
resnet18=torchvision.models.resnet18(pretrained=True)
print(resnet18)
ResNet(
(conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
(layer1): Sequential(
(0): BasicBlock(
(conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(1): BasicBlock(
(conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(layer2): Sequential(
(0): BasicBlock(
(conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(downsample): Sequential(
(0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): BasicBlock(
(conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(layer3): Sequential(
(0): BasicBlock(
(conv1): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(downsample): Sequential(
(0): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): BasicBlock(
(conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(layer4): Sequential(
(0): BasicBlock(
(conv1): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(downsample): Sequential(
(0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): BasicBlock(
(conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(avgpool): AdaptiveAvgPool2d(output_size=(1, 1))
(fc): Linear(in_features=512, out_features=1000, bias=True)
)
### Plot accuracy and loss
from bokeh.io import output_file, show
from bokeh.models import ColumnDataSource, FactorRange
from bokeh.plotting import figure
from bokeh.plotting import output_notebook
from bokeh.transform import dodge
output_notebook()
def plot_history(source,title,epoch_num):
# TODO: Plot training and validation curves for avg. loss and accuracy
p = figure(x_range=epoch_num, plot_height=400, title='Train '+title+' vs Test '+title,
toolbar_location=None, tools="")
#p.vbar(x='x', top='counts', width=0.9, source=source, fill_color=factor_cmap('x', palette=palette, factors=years, start=1, end=2))
p.vbar(x=dodge('epoch_num', -0.25, range=p.x_range), top='Train', width=0.2, source=source,
color="#B4EEB4", legend_label="Train")
p.vbar(x=dodge('epoch_num', 0.0, range=p.x_range), top='Validation', width=0.2, source=source,
color="#FDD6E5", legend_label="Validation")
p.x_range.range_padding = 0.1
p.xgrid.grid_line_color = None
p.legend.location = "top_left"
p.legend.orientation = "horizontal"
p.xaxis.axis_label = "Epoch Number"
p.yaxis.axis_label = title
show(p)
### Confusion Matrix and Classification Report
import seaborn as sn
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.metrics import classification_report
def confusion_matrix_classification_report(network, test_iter, num_classes=100, device='cuda'):
network.eval()
mat = torch.zeros(size=(num_classes, num_classes), dtype=torch.get_default_dtype(),
device='cpu')
y_true=[]
y_pred=[]
for data,label in test_iter:
data = data.to(device)
result = network(data)
result = torch.argmax(result, dim=-1).data.cpu()
idx = torch.stack([label, result], dim=1)
#print(idx)
idx, count = torch.unique(idx, dim=0, return_counts=True)
mat[idx[:,0], idx[:,1]] += count
y_true.append(label.cpu().numpy())
y_pred.append(result.numpy())
flat_pred = []
flat_true = []
for i in range(len(y_pred)):
for j in range(len(y_pred[i])):
flat_pred.append(y_pred[i][j])
flat_true.append(y_true[i][j])
#print("number of testing samples results:", len(flat_pred))
class_report=classification_report(flat_true,flat_pred)
mat /= torch.sum(mat, dim=1, keepdims=True)
return mat,class_report
### Method to freeze layers as required
### i=0 Train using all layers
### i=1 freeze all layers except the first one
### i=2 freeze all layers except the second one
### i=3 freeze all layers except the third one
### i=4 freeze all layers except the fourth one
def freeze_resnet(i):
resnet18=torchvision.models.resnet18(pretrained=True)
if(i!=0): #### Freeze all layers except layer i
for name, child in resnet18.named_children():
if(name!='layer'+'i'):
for name2, params in child.named_parameters():
params.requires_grad = False
resnet18=nn.Sequential(resnet18,nn.Linear(1000, 100))
resnet18=resnet18.to(device)
return resnet18
for i in range(5):
if(i!=0):
title=" Train Resnet by freezing All layers except Layer : "+str(i)+" "
else:
title=" Train Resnet by All layers "
print("***************================="+title+"===================***************")
resnet18=freeze_resnet(i)
criterion = nn.CrossEntropyLoss()
train_loss,train_accuracy,validation_loss,validation_accuracy=start_train_valid(resnet18,train_dataloader,validation_dataloader,criterion)
epoch_num = []
for i in range(num_epochs):
epoch_num.append(str(i+1))
source_loss = ColumnDataSource(data={'epoch_num': epoch_num,'Train': train_loss,'Validation': validation_loss })
plot_history(source_loss,'Loss',epoch_num)
source_acc = ColumnDataSource(data={'epoch_num': epoch_num,'Train': train_accuracy,'Validation': validation_accuracy })
plot_history(source_acc,'Accuracy',epoch_num)
result_matrix,class_report = confusion_matrix_classification_report(resnet18,test_dataloader,device=device)
print("\nConfusion Matrix : \n",result_matrix)
print("\nClassification Report : \n",class_report)
plot_matrix = pd.DataFrame(result_matrix.numpy())
plt.figure(figsize = (25,25))
sn.heatmap(plot_matrix, annot=True, cmap="YlGnBu")
plt.title(title)
plt.show()
print("\n***************=================================================================***************\n\n")
***************================= Train Resnet by All layers ===================*************** => Training Epoch [ 1/ 10] Training Loss: 4.8381% Training Accuracy: 2.948% | Validation Epoch #1 Validation Loss: 26.1114% Validation Accuracy: 2.64% | Elapsed time : 0:00:37 => Training Epoch [ 2/ 10] Training Loss: 3.9678% Training Accuracy: 7.760% | Validation Epoch #2 Validation Loss: 3.7023% Validation Accuracy: 10.96% | Elapsed time : 0:01:16 => Training Epoch [ 3/ 10] Training Loss: 3.4640% Training Accuracy: 15.764% | Validation Epoch #3 Validation Loss: 3.3565% Validation Accuracy: 18.15% | Elapsed time : 0:01:54 => Training Epoch [ 4/ 10] Training Loss: 3.0140% Training Accuracy: 24.104% | Validation Epoch #4 Validation Loss: 2.9785% Validation Accuracy: 25.86% | Elapsed time : 0:02:33 => Training Epoch [ 5/ 10] Training Loss: 2.6886% Training Accuracy: 30.531% | Validation Epoch #5 Validation Loss: 2.7744% Validation Accuracy: 30.04% | Elapsed time : 0:03:12 => Training Epoch [ 6/ 10] Training Loss: 2.4545% Training Accuracy: 35.619% | Validation Epoch #6 Validation Loss: 2.7284% Validation Accuracy: 32.51% | Elapsed time : 0:03:51 => Training Epoch [ 7/ 10] Training Loss: 2.2932% Training Accuracy: 38.988% | Validation Epoch #7 Validation Loss: 2.6171% Validation Accuracy: 33.23% | Elapsed time : 0:04:29 => Training Epoch [ 8/ 10] Training Loss: 2.1269% Training Accuracy: 42.680% | Validation Epoch #8 Validation Loss: 2.5256% Validation Accuracy: 36.23% | Elapsed time : 0:05:08 => Training Epoch [ 9/ 10] Training Loss: 1.9856% Training Accuracy: 46.007% | Validation Epoch #9 Validation Loss: 3.0575% Validation Accuracy: 35.77% | Elapsed time : 0:05:48 => Training Epoch [ 10/ 10] Training Loss: 1.8426% Training Accuracy: 49.351% | Validation Epoch #10 Validation Loss: 3.7367% Validation Accuracy: 21.51% | Elapsed time : 0:06:27
Confusion Matrix :
tensor([[0.2900, 0.0000, 0.0100, ..., 0.0000, 0.0000, 0.0000],
[0.0100, 0.3000, 0.0300, ..., 0.0000, 0.0000, 0.0300],
[0.0000, 0.0000, 0.3300, ..., 0.0100, 0.0400, 0.0100],
...,
[0.0000, 0.0000, 0.0100, ..., 0.1800, 0.0000, 0.0000],
[0.0000, 0.0000, 0.1600, ..., 0.0200, 0.1000, 0.0200],
[0.0100, 0.0000, 0.0100, ..., 0.0100, 0.0000, 0.2500]])
Classification Report :
precision recall f1-score support
0 0.64 0.29 0.40 100
1 0.38 0.30 0.33 100
2 0.12 0.33 0.18 100
3 0.00 0.00 0.00 100
4 0.20 0.02 0.04 100
5 0.15 0.19 0.17 100
6 0.31 0.41 0.35 100
7 0.38 0.33 0.35 100
8 0.11 0.74 0.19 100
9 0.17 0.49 0.25 100
10 0.36 0.04 0.07 100
11 0.30 0.09 0.14 100
12 0.19 0.25 0.21 100
13 0.26 0.10 0.14 100
14 0.26 0.22 0.24 100
15 0.18 0.09 0.12 100
16 0.18 0.23 0.20 100
17 0.71 0.20 0.31 100
18 0.17 0.08 0.11 100
19 0.15 0.20 0.17 100
20 0.07 0.55 0.12 100
21 0.63 0.12 0.20 100
22 0.16 0.30 0.21 100
23 0.81 0.13 0.22 100
24 0.25 0.01 0.02 100
25 0.12 0.14 0.13 100
26 0.30 0.16 0.21 100
27 0.22 0.04 0.07 100
28 0.46 0.36 0.40 100
29 0.39 0.07 0.12 100
30 0.43 0.26 0.33 100
31 0.24 0.32 0.27 100
32 0.40 0.08 0.13 100
33 0.33 0.37 0.35 100
34 0.17 0.19 0.18 100
35 0.38 0.05 0.09 100
36 0.60 0.09 0.16 100
37 0.31 0.19 0.24 100
38 0.16 0.26 0.20 100
39 0.48 0.33 0.39 100
40 0.23 0.29 0.25 100
41 0.64 0.23 0.34 100
42 0.09 0.53 0.16 100
43 0.32 0.19 0.24 100
44 0.03 0.01 0.01 100
45 0.07 0.09 0.08 100
46 0.32 0.15 0.20 100
47 0.35 0.25 0.29 100
48 0.24 0.33 0.28 100
49 0.80 0.12 0.21 100
50 0.09 0.05 0.06 100
51 0.08 0.07 0.08 100
52 0.62 0.13 0.21 100
53 0.47 0.35 0.40 100
54 0.49 0.37 0.42 100
55 0.00 0.00 0.00 100
56 0.68 0.46 0.55 100
57 0.71 0.10 0.18 100
58 0.77 0.10 0.18 100
59 0.28 0.33 0.30 100
60 0.57 0.63 0.60 100
61 0.38 0.21 0.27 100
62 0.23 0.54 0.32 100
63 0.11 0.32 0.16 100
64 0.17 0.07 0.10 100
65 0.07 0.23 0.11 100
66 0.22 0.31 0.26 100
67 0.23 0.06 0.10 100
68 0.78 0.39 0.52 100
69 0.51 0.26 0.34 100
70 0.31 0.53 0.39 100
71 0.49 0.39 0.43 100
72 0.00 0.00 0.00 100
73 0.32 0.11 0.16 100
74 0.18 0.09 0.12 100
75 0.95 0.18 0.30 100
76 0.46 0.63 0.53 100
77 0.11 0.02 0.03 100
78 0.20 0.13 0.16 100
79 0.19 0.38 0.26 100
80 0.15 0.10 0.12 100
81 0.57 0.04 0.07 100
82 0.46 0.40 0.43 100
83 0.29 0.13 0.18 100
84 0.08 0.29 0.13 100
85 0.67 0.04 0.08 100
86 0.17 0.17 0.17 100
87 0.12 0.18 0.14 100
88 0.16 0.36 0.22 100
89 0.24 0.08 0.12 100
90 0.60 0.03 0.06 100
91 0.62 0.13 0.21 100
92 0.31 0.11 0.16 100
93 0.18 0.04 0.07 100
94 0.45 0.71 0.55 100
95 0.49 0.28 0.36 100
96 0.42 0.15 0.22 100
97 0.32 0.18 0.23 100
98 0.26 0.10 0.14 100
99 0.14 0.25 0.18 100
accuracy 0.22 10000
macro avg 0.33 0.22 0.21 10000
weighted avg 0.33 0.22 0.21 10000
***************=================================================================*************** ***************================= Train Resnet by freezing All layers except Layer : 1 ===================*************** => Training Epoch [ 1/ 10] Training Loss: 24.1925% Training Accuracy: 11.895% | Validation Epoch #1 Validation Loss: 28.4622% Validation Accuracy: 13.15% | Elapsed time : 0:00:22 => Training Epoch [ 2/ 10] Training Loss: 26.3396% Training Accuracy: 15.078% | Validation Epoch #2 Validation Loss: 28.5056% Validation Accuracy: 14.75% | Elapsed time : 0:00:46 => Training Epoch [ 3/ 10] Training Loss: 26.1546% Training Accuracy: 15.563% | Validation Epoch #3 Validation Loss: 29.6301% Validation Accuracy: 16.04% | Elapsed time : 0:01:09 => Training Epoch [ 4/ 10] Training Loss: 26.3216% Training Accuracy: 16.675% | Validation Epoch #4 Validation Loss: 31.1082% Validation Accuracy: 15.34% | Elapsed time : 0:01:32 => Training Epoch [ 5/ 10] Training Loss: 26.3413% Training Accuracy: 17.036% | Validation Epoch #5 Validation Loss: 28.9149% Validation Accuracy: 15.58% | Elapsed time : 0:01:56 => Training Epoch [ 6/ 10] Training Loss: 25.8228% Training Accuracy: 17.411% | Validation Epoch #6 Validation Loss: 31.7716% Validation Accuracy: 15.19% | Elapsed time : 0:02:18 => Training Epoch [ 7/ 10] Training Loss: 26.2115% Training Accuracy: 17.645% | Validation Epoch #7 Validation Loss: 30.9946% Validation Accuracy: 14.79% | Elapsed time : 0:02:40 => Training Epoch [ 8/ 10] Training Loss: 26.1405% Training Accuracy: 17.792% | Validation Epoch #8 Validation Loss: 30.4202% Validation Accuracy: 14.83% | Elapsed time : 0:03:03 => Training Epoch [ 9/ 10] Training Loss: 25.7810% Training Accuracy: 18.427% | Validation Epoch #9 Validation Loss: 31.3585% Validation Accuracy: 15.55% | Elapsed time : 0:03:25 => Training Epoch [ 10/ 10] Training Loss: 26.1643% Training Accuracy: 18.195% | Validation Epoch #10 Validation Loss: 30.3724% Validation Accuracy: 16.03% | Elapsed time : 0:03:48
Confusion Matrix :
tensor([[0.6700, 0.0100, 0.0000, ..., 0.0000, 0.0000, 0.0000],
[0.0000, 0.2400, 0.0300, ..., 0.0000, 0.0000, 0.0000],
[0.0200, 0.0400, 0.0700, ..., 0.0100, 0.0100, 0.0100],
...,
[0.0000, 0.0300, 0.0000, ..., 0.1300, 0.0200, 0.0000],
[0.0000, 0.0100, 0.0600, ..., 0.0000, 0.0100, 0.0000],
[0.0100, 0.0200, 0.0000, ..., 0.0000, 0.0000, 0.3100]])
Classification Report :
precision recall f1-score support
0 0.72 0.67 0.69 100
1 0.14 0.24 0.17 100
2 0.11 0.07 0.08 100
3 0.06 0.09 0.07 100
4 0.03 0.03 0.03 100
5 0.12 0.09 0.10 100
6 0.09 0.18 0.12 100
7 0.23 0.26 0.25 100
8 0.14 0.15 0.15 100
9 0.14 0.28 0.19 100
10 0.24 0.13 0.17 100
11 0.00 0.00 0.00 100
12 0.13 0.12 0.12 100
13 0.06 0.06 0.06 100
14 0.04 0.01 0.02 100
15 0.07 0.06 0.06 100
16 0.25 0.16 0.19 100
17 0.20 0.12 0.15 100
18 0.20 0.16 0.18 100
19 0.08 0.11 0.10 100
20 0.21 0.36 0.27 100
21 0.20 0.15 0.17 100
22 0.25 0.02 0.04 100
23 0.22 0.39 0.28 100
24 0.25 0.24 0.24 100
25 0.09 0.16 0.11 100
26 0.08 0.19 0.12 100
27 0.10 0.02 0.03 100
28 0.16 0.23 0.19 100
29 0.06 0.03 0.04 100
30 0.20 0.31 0.24 100
31 0.15 0.13 0.14 100
32 0.07 0.08 0.08 100
33 0.14 0.17 0.15 100
34 0.06 0.09 0.07 100
35 0.10 0.08 0.09 100
36 0.11 0.12 0.11 100
37 0.06 0.02 0.03 100
38 0.00 0.00 0.00 100
39 0.34 0.76 0.47 100
40 0.17 0.02 0.04 100
41 0.23 0.25 0.24 100
42 0.09 0.04 0.06 100
43 0.18 0.14 0.16 100
44 0.02 0.01 0.01 100
45 0.02 0.01 0.01 100
46 0.09 0.16 0.11 100
47 0.09 0.23 0.13 100
48 0.44 0.31 0.36 100
49 0.44 0.08 0.14 100
50 0.06 0.03 0.04 100
51 0.05 0.02 0.03 100
52 0.32 0.27 0.29 100
53 0.54 0.63 0.58 100
54 0.24 0.34 0.28 100
55 0.03 0.04 0.03 100
56 0.23 0.24 0.24 100
57 0.14 0.15 0.14 100
58 0.09 0.13 0.11 100
59 0.14 0.20 0.17 100
60 0.74 0.28 0.41 100
61 0.34 0.39 0.36 100
62 0.25 0.28 0.26 100
63 0.19 0.15 0.17 100
64 0.05 0.07 0.06 100
65 0.06 0.03 0.04 100
66 0.08 0.01 0.02 100
67 0.29 0.07 0.11 100
68 0.40 0.25 0.31 100
69 0.18 0.29 0.22 100
70 0.33 0.27 0.30 100
71 0.33 0.31 0.32 100
72 0.06 0.04 0.05 100
73 0.34 0.10 0.16 100
74 0.07 0.09 0.08 100
75 0.31 0.20 0.24 100
76 0.42 0.32 0.36 100
77 0.10 0.06 0.07 100
78 0.09 0.12 0.10 100
79 0.11 0.16 0.13 100
80 0.00 0.00 0.00 100
81 0.00 0.00 0.00 100
82 0.49 0.60 0.54 100
83 0.17 0.26 0.21 100
84 0.07 0.03 0.04 100
85 0.09 0.08 0.08 100
86 0.12 0.27 0.16 100
87 0.29 0.49 0.37 100
88 0.10 0.08 0.09 100
89 0.12 0.14 0.13 100
90 0.09 0.11 0.10 100
91 0.04 0.03 0.03 100
92 0.19 0.18 0.18 100
93 0.07 0.17 0.10 100
94 0.50 0.62 0.55 100
95 0.32 0.08 0.13 100
96 0.26 0.21 0.23 100
97 0.16 0.13 0.14 100
98 0.02 0.01 0.01 100
99 0.27 0.31 0.29 100
accuracy 0.17 10000
macro avg 0.18 0.17 0.16 10000
weighted avg 0.18 0.17 0.16 10000
***************=================================================================*************** ***************================= Train Resnet by freezing All layers except Layer : 2 ===================*************** => Training Epoch [ 1/ 10] Training Loss: 24.6535% Training Accuracy: 11.913% | Validation Epoch #1 Validation Loss: 29.2039% Validation Accuracy: 13.30% | Elapsed time : 0:00:23 => Training Epoch [ 2/ 10] Training Loss: 26.2579% Training Accuracy: 15.155% | Validation Epoch #2 Validation Loss: 29.4709% Validation Accuracy: 14.38% | Elapsed time : 0:00:45 => Training Epoch [ 3/ 10] Training Loss: 26.1884% Training Accuracy: 15.866% | Validation Epoch #3 Validation Loss: 28.7980% Validation Accuracy: 15.52% | Elapsed time : 0:01:08 => Training Epoch [ 4/ 10] Training Loss: 25.8310% Training Accuracy: 16.934% | Validation Epoch #4 Validation Loss: 29.8791% Validation Accuracy: 14.86% | Elapsed time : 0:01:30 => Training Epoch [ 5/ 10] Training Loss: 25.7989% Training Accuracy: 17.046% | Validation Epoch #5 Validation Loss: 30.0134% Validation Accuracy: 14.79% | Elapsed time : 0:01:54 => Training Epoch [ 6/ 10] Training Loss: 26.3427% Training Accuracy: 16.966% | Validation Epoch #6 Validation Loss: 29.6805% Validation Accuracy: 15.45% | Elapsed time : 0:02:16 => Training Epoch [ 7/ 10] Training Loss: 26.0961% Training Accuracy: 17.543% | Validation Epoch #7 Validation Loss: 30.9269% Validation Accuracy: 15.04% | Elapsed time : 0:02:39 => Training Epoch [ 8/ 10] Training Loss: 26.0951% Training Accuracy: 17.802% | Validation Epoch #8 Validation Loss: 29.3633% Validation Accuracy: 16.71% | Elapsed time : 0:03:02 => Training Epoch [ 9/ 10] Training Loss: 25.6515% Training Accuracy: 18.198% | Validation Epoch #9 Validation Loss: 30.3421% Validation Accuracy: 16.29% | Elapsed time : 0:03:25 => Training Epoch [ 10/ 10] Training Loss: 25.6180% Training Accuracy: 18.250% | Validation Epoch #10 Validation Loss: 32.3724% Validation Accuracy: 15.00% | Elapsed time : 0:03:48
Confusion Matrix :
tensor([[0.7100, 0.0100, 0.0000, ..., 0.0000, 0.0000, 0.0000],
[0.0100, 0.1700, 0.0000, ..., 0.0000, 0.0000, 0.0000],
[0.0000, 0.0000, 0.0900, ..., 0.0000, 0.0300, 0.0000],
...,
[0.0000, 0.0100, 0.0000, ..., 0.1400, 0.0400, 0.0000],
[0.0000, 0.0000, 0.0300, ..., 0.0000, 0.0600, 0.0000],
[0.0100, 0.0100, 0.0000, ..., 0.0000, 0.0100, 0.0700]])
Classification Report :
precision recall f1-score support
0 0.53 0.71 0.61 100
1 0.22 0.17 0.19 100
2 0.19 0.09 0.12 100
3 0.13 0.02 0.03 100
4 0.06 0.01 0.02 100
5 0.19 0.04 0.07 100
6 0.10 0.18 0.13 100
7 0.20 0.11 0.14 100
8 0.07 0.19 0.11 100
9 0.14 0.29 0.19 100
10 0.19 0.25 0.21 100
11 0.03 0.02 0.02 100
12 0.20 0.09 0.12 100
13 0.13 0.15 0.14 100
14 0.11 0.12 0.12 100
15 0.09 0.04 0.05 100
16 0.26 0.20 0.22 100
17 0.15 0.18 0.17 100
18 0.15 0.11 0.13 100
19 0.01 0.01 0.01 100
20 0.22 0.24 0.23 100
21 0.08 0.09 0.09 100
22 0.26 0.25 0.26 100
23 0.29 0.20 0.24 100
24 0.35 0.19 0.25 100
25 0.11 0.15 0.13 100
26 0.08 0.24 0.12 100
27 0.09 0.03 0.04 100
28 0.15 0.05 0.08 100
29 0.05 0.08 0.06 100
30 0.15 0.55 0.24 100
31 0.13 0.04 0.06 100
32 0.08 0.09 0.08 100
33 0.17 0.16 0.16 100
34 0.06 0.06 0.06 100
35 0.06 0.09 0.07 100
36 0.09 0.09 0.09 100
37 0.11 0.06 0.08 100
38 0.05 0.04 0.04 100
39 0.71 0.29 0.41 100
40 0.09 0.10 0.09 100
41 0.26 0.23 0.24 100
42 0.12 0.08 0.10 100
43 0.16 0.18 0.17 100
44 0.07 0.06 0.06 100
45 0.06 0.13 0.08 100
46 0.13 0.19 0.15 100
47 0.15 0.16 0.16 100
48 0.27 0.49 0.35 100
49 0.16 0.23 0.19 100
50 0.02 0.01 0.01 100
51 0.06 0.07 0.06 100
52 0.30 0.21 0.25 100
53 0.63 0.36 0.46 100
54 0.24 0.23 0.24 100
55 0.04 0.05 0.05 100
56 0.07 0.08 0.08 100
57 0.26 0.06 0.10 100
58 0.07 0.11 0.08 100
59 0.09 0.15 0.11 100
60 0.60 0.28 0.38 100
61 0.25 0.16 0.20 100
62 0.33 0.03 0.06 100
63 0.08 0.19 0.11 100
64 0.06 0.02 0.03 100
65 0.06 0.05 0.06 100
66 0.08 0.09 0.09 100
67 0.16 0.12 0.14 100
68 0.31 0.19 0.23 100
69 0.15 0.12 0.13 100
70 0.26 0.26 0.26 100
71 0.34 0.16 0.22 100
72 0.04 0.02 0.03 100
73 0.11 0.17 0.13 100
74 0.06 0.09 0.07 100
75 0.21 0.22 0.22 100
76 0.28 0.22 0.24 100
77 0.07 0.06 0.06 100
78 0.17 0.05 0.08 100
79 0.06 0.03 0.04 100
80 0.04 0.08 0.05 100
81 0.17 0.06 0.09 100
82 0.37 0.60 0.46 100
83 0.28 0.05 0.08 100
84 0.03 0.04 0.03 100
85 0.07 0.05 0.06 100
86 0.24 0.18 0.21 100
87 0.24 0.48 0.32 100
88 0.04 0.02 0.03 100
89 0.20 0.07 0.10 100
90 0.08 0.05 0.06 100
91 0.16 0.09 0.12 100
92 0.25 0.14 0.18 100
93 0.06 0.02 0.03 100
94 0.29 0.73 0.41 100
95 0.16 0.04 0.06 100
96 0.18 0.31 0.23 100
97 0.11 0.14 0.12 100
98 0.05 0.06 0.06 100
99 0.30 0.07 0.11 100
accuracy 0.15 10000
macro avg 0.17 0.15 0.14 10000
weighted avg 0.17 0.15 0.14 10000
***************=================================================================*************** ***************================= Train Resnet by freezing All layers except Layer : 3 ===================*************** => Training Epoch [ 1/ 10] Training Loss: 24.7298% Training Accuracy: 11.980% | Validation Epoch #1 Validation Loss: 27.0506% Validation Accuracy: 14.85% | Elapsed time : 0:00:22 => Training Epoch [ 2/ 10] Training Loss: 26.3978% Training Accuracy: 14.993% | Validation Epoch #2 Validation Loss: 30.7062% Validation Accuracy: 14.42% | Elapsed time : 0:00:45 => Training Epoch [ 3/ 10] Training Loss: 26.4328% Training Accuracy: 16.088% | Validation Epoch #3 Validation Loss: 29.6265% Validation Accuracy: 14.50% | Elapsed time : 0:01:09 => Training Epoch [ 4/ 10] Training Loss: 26.1429% Training Accuracy: 16.789% | Validation Epoch #4 Validation Loss: 29.6214% Validation Accuracy: 15.77% | Elapsed time : 0:01:32 => Training Epoch [ 5/ 10] Training Loss: 26.0662% Training Accuracy: 16.983% | Validation Epoch #5 Validation Loss: 29.3589% Validation Accuracy: 16.12% | Elapsed time : 0:01:53 => Training Epoch [ 6/ 10] Training Loss: 26.0073% Training Accuracy: 17.247% | Validation Epoch #6 Validation Loss: 29.7176% Validation Accuracy: 15.86% | Elapsed time : 0:02:15 => Training Epoch [ 7/ 10] Training Loss: 25.8890% Training Accuracy: 17.623% | Validation Epoch #7 Validation Loss: 30.2360% Validation Accuracy: 16.40% | Elapsed time : 0:02:37 => Training Epoch [ 8/ 10] Training Loss: 25.8285% Training Accuracy: 17.912% | Validation Epoch #8 Validation Loss: 29.4937% Validation Accuracy: 15.75% | Elapsed time : 0:03:00 => Training Epoch [ 9/ 10] Training Loss: 26.0967% Training Accuracy: 17.872% | Validation Epoch #9 Validation Loss: 29.5462% Validation Accuracy: 15.78% | Elapsed time : 0:03:22 => Training Epoch [ 10/ 10] Training Loss: 25.6660% Training Accuracy: 18.168% | Validation Epoch #10 Validation Loss: 29.8650% Validation Accuracy: 15.48% | Elapsed time : 0:03:44
Confusion Matrix :
tensor([[0.5900, 0.0100, 0.0000, ..., 0.0000, 0.0100, 0.0000],
[0.0100, 0.0900, 0.0000, ..., 0.0000, 0.0000, 0.0000],
[0.0100, 0.0000, 0.0200, ..., 0.0000, 0.0200, 0.0200],
...,
[0.0000, 0.0000, 0.0000, ..., 0.1200, 0.0000, 0.0000],
[0.0000, 0.0100, 0.0000, ..., 0.0000, 0.1000, 0.0000],
[0.0000, 0.0000, 0.0000, ..., 0.0000, 0.0000, 0.2800]])
Classification Report :
precision recall f1-score support
0 0.87 0.59 0.70 100
1 0.11 0.09 0.10 100
2 0.11 0.02 0.03 100
3 0.04 0.08 0.06 100
4 0.02 0.07 0.03 100
5 0.05 0.04 0.04 100
6 0.12 0.17 0.14 100
7 0.21 0.16 0.18 100
8 0.19 0.21 0.20 100
9 0.19 0.34 0.24 100
10 0.38 0.08 0.13 100
11 0.06 0.18 0.09 100
12 0.11 0.03 0.05 100
13 0.15 0.13 0.14 100
14 0.10 0.08 0.09 100
15 0.06 0.17 0.09 100
16 0.33 0.22 0.26 100
17 0.17 0.05 0.08 100
18 0.19 0.06 0.09 100
19 0.07 0.05 0.06 100
20 0.47 0.18 0.26 100
21 0.20 0.06 0.09 100
22 0.21 0.28 0.24 100
23 0.21 0.24 0.22 100
24 0.38 0.35 0.36 100
25 0.12 0.13 0.13 100
26 0.13 0.08 0.10 100
27 0.07 0.03 0.04 100
28 0.05 0.01 0.02 100
29 0.16 0.07 0.10 100
30 0.09 0.04 0.05 100
31 0.10 0.07 0.08 100
32 0.05 0.05 0.05 100
33 0.28 0.19 0.23 100
34 0.09 0.20 0.12 100
35 0.05 0.05 0.05 100
36 0.10 0.03 0.05 100
37 0.05 0.11 0.07 100
38 0.06 0.03 0.04 100
39 0.59 0.41 0.48 100
40 0.05 0.04 0.05 100
41 0.54 0.30 0.38 100
42 0.19 0.10 0.13 100
43 0.12 0.13 0.13 100
44 0.00 0.00 0.00 100
45 0.13 0.03 0.05 100
46 0.11 0.16 0.13 100
47 0.35 0.09 0.14 100
48 0.19 0.33 0.24 100
49 0.17 0.46 0.25 100
50 0.02 0.01 0.01 100
51 0.06 0.07 0.06 100
52 0.35 0.40 0.38 100
53 0.45 0.79 0.57 100
54 0.36 0.27 0.31 100
55 0.03 0.02 0.03 100
56 0.28 0.21 0.24 100
57 0.28 0.24 0.26 100
58 0.12 0.06 0.08 100
59 0.13 0.26 0.17 100
60 0.49 0.35 0.41 100
61 0.38 0.50 0.43 100
62 0.41 0.11 0.17 100
63 0.11 0.15 0.13 100
64 0.03 0.02 0.02 100
65 0.03 0.01 0.01 100
66 0.09 0.08 0.08 100
67 0.12 0.22 0.15 100
68 0.39 0.23 0.29 100
69 0.40 0.25 0.31 100
70 0.22 0.17 0.19 100
71 0.17 0.22 0.19 100
72 0.03 0.04 0.04 100
73 0.13 0.08 0.10 100
74 0.06 0.08 0.07 100
75 0.31 0.15 0.20 100
76 0.41 0.22 0.29 100
77 0.06 0.09 0.07 100
78 0.08 0.13 0.10 100
79 0.04 0.01 0.02 100
80 0.02 0.03 0.02 100
81 0.13 0.10 0.11 100
82 0.55 0.53 0.54 100
83 0.12 0.05 0.07 100
84 0.05 0.05 0.05 100
85 0.08 0.24 0.12 100
86 0.59 0.23 0.33 100
87 0.25 0.54 0.34 100
88 0.04 0.07 0.05 100
89 0.09 0.11 0.10 100
90 0.09 0.14 0.11 100
91 0.09 0.15 0.11 100
92 0.19 0.26 0.22 100
93 0.09 0.07 0.08 100
94 0.65 0.40 0.49 100
95 0.22 0.17 0.19 100
96 0.24 0.06 0.10 100
97 0.13 0.12 0.13 100
98 0.12 0.10 0.11 100
99 0.29 0.28 0.28 100
accuracy 0.16 10000
macro avg 0.19 0.16 0.16 10000
weighted avg 0.19 0.16 0.16 10000
***************=================================================================*************** ***************================= Train Resnet by freezing All layers except Layer : 4 ===================*************** => Training Epoch [ 1/ 10] Training Loss: 25.0883% Training Accuracy: 11.796% | Validation Epoch #1 Validation Loss: 28.3594% Validation Accuracy: 14.76% | Elapsed time : 0:00:21 => Training Epoch [ 2/ 10] Training Loss: 26.6855% Training Accuracy: 14.943% | Validation Epoch #2 Validation Loss: 28.1927% Validation Accuracy: 14.80% | Elapsed time : 0:00:44 => Training Epoch [ 3/ 10] Training Loss: 26.2319% Training Accuracy: 16.036% | Validation Epoch #3 Validation Loss: 30.6342% Validation Accuracy: 14.68% | Elapsed time : 0:01:05 => Training Epoch [ 4/ 10] Training Loss: 25.9204% Training Accuracy: 16.533% | Validation Epoch #4 Validation Loss: 30.5431% Validation Accuracy: 15.46% | Elapsed time : 0:01:26 => Training Epoch [ 5/ 10] Training Loss: 26.1958% Training Accuracy: 16.707% | Validation Epoch #5 Validation Loss: 30.7355% Validation Accuracy: 15.55% | Elapsed time : 0:01:49 => Training Epoch [ 6/ 10] Training Loss: 25.8954% Training Accuracy: 17.563% | Validation Epoch #6 Validation Loss: 29.5973% Validation Accuracy: 16.11% | Elapsed time : 0:02:11 => Training Epoch [ 7/ 10] Training Loss: 26.2763% Training Accuracy: 17.605% | Validation Epoch #7 Validation Loss: 31.0453% Validation Accuracy: 15.06% | Elapsed time : 0:02:33 => Training Epoch [ 8/ 10] Training Loss: 26.0500% Training Accuracy: 17.650% | Validation Epoch #8 Validation Loss: 28.9701% Validation Accuracy: 16.46% | Elapsed time : 0:02:55 => Training Epoch [ 9/ 10] Training Loss: 25.8172% Training Accuracy: 18.183% | Validation Epoch #9 Validation Loss: 30.2980% Validation Accuracy: 15.04% | Elapsed time : 0:03:19 => Training Epoch [ 10/ 10] Training Loss: 25.6745% Training Accuracy: 18.230% | Validation Epoch #10 Validation Loss: 30.1029% Validation Accuracy: 14.85% | Elapsed time : 0:03:41
Confusion Matrix :
tensor([[0.5300, 0.0000, 0.0100, ..., 0.0000, 0.0000, 0.0000],
[0.0000, 0.0400, 0.0300, ..., 0.0000, 0.0200, 0.0000],
[0.0400, 0.0000, 0.2000, ..., 0.0000, 0.0900, 0.0100],
...,
[0.0000, 0.0000, 0.0400, ..., 0.0700, 0.0200, 0.0000],
[0.0000, 0.0000, 0.0500, ..., 0.0000, 0.1600, 0.0000],
[0.0100, 0.0000, 0.0000, ..., 0.0000, 0.0200, 0.3300]])
Classification Report :
precision recall f1-score support
0 0.60 0.53 0.56 100
1 0.11 0.04 0.06 100
2 0.07 0.20 0.11 100
3 0.10 0.03 0.05 100
4 0.04 0.03 0.04 100
5 0.08 0.08 0.08 100
6 0.07 0.21 0.11 100
7 0.43 0.18 0.25 100
8 0.26 0.23 0.24 100
9 0.12 0.06 0.08 100
10 0.19 0.06 0.09 100
11 0.06 0.02 0.03 100
12 0.16 0.08 0.11 100
13 0.06 0.04 0.05 100
14 0.15 0.14 0.14 100
15 0.05 0.03 0.04 100
16 0.35 0.11 0.17 100
17 0.23 0.14 0.18 100
18 0.22 0.27 0.24 100
19 0.08 0.08 0.08 100
20 0.24 0.51 0.32 100
21 0.10 0.09 0.10 100
22 0.20 0.16 0.18 100
23 0.35 0.06 0.10 100
24 0.23 0.18 0.20 100
25 0.01 0.01 0.01 100
26 0.11 0.09 0.10 100
27 0.10 0.07 0.08 100
28 0.11 0.16 0.13 100
29 0.12 0.04 0.06 100
30 0.14 0.11 0.12 100
31 0.16 0.15 0.15 100
32 0.05 0.09 0.06 100
33 0.34 0.19 0.24 100
34 0.06 0.13 0.08 100
35 0.14 0.05 0.07 100
36 0.08 0.03 0.04 100
37 0.09 0.13 0.10 100
38 0.03 0.03 0.03 100
39 0.57 0.64 0.60 100
40 0.07 0.10 0.08 100
41 0.29 0.36 0.32 100
42 0.15 0.04 0.06 100
43 0.17 0.20 0.18 100
44 0.00 0.00 0.00 100
45 0.03 0.02 0.02 100
46 0.06 0.04 0.05 100
47 0.33 0.26 0.29 100
48 0.32 0.33 0.33 100
49 0.22 0.41 0.29 100
50 0.08 0.07 0.07 100
51 0.06 0.06 0.06 100
52 0.39 0.30 0.34 100
53 0.65 0.55 0.59 100
54 0.20 0.11 0.14 100
55 0.02 0.01 0.01 100
56 0.16 0.34 0.22 100
57 0.14 0.28 0.19 100
58 0.07 0.11 0.08 100
59 0.13 0.16 0.14 100
60 0.43 0.38 0.40 100
61 0.59 0.16 0.25 100
62 0.21 0.31 0.25 100
63 0.14 0.09 0.11 100
64 0.06 0.12 0.08 100
65 0.05 0.05 0.05 100
66 0.03 0.02 0.02 100
67 0.26 0.22 0.24 100
68 0.15 0.06 0.09 100
69 0.27 0.28 0.28 100
70 0.43 0.12 0.19 100
71 0.34 0.10 0.16 100
72 0.00 0.00 0.00 100
73 0.08 0.19 0.11 100
74 0.07 0.06 0.06 100
75 0.16 0.14 0.15 100
76 0.49 0.28 0.36 100
77 0.07 0.18 0.10 100
78 0.16 0.13 0.14 100
79 0.14 0.13 0.13 100
80 0.04 0.10 0.05 100
81 0.08 0.04 0.05 100
82 0.60 0.42 0.49 100
83 0.12 0.31 0.17 100
84 0.25 0.02 0.04 100
85 0.09 0.12 0.10 100
86 0.51 0.23 0.32 100
87 0.28 0.48 0.36 100
88 0.14 0.09 0.11 100
89 0.06 0.09 0.07 100
90 0.10 0.09 0.09 100
91 0.09 0.02 0.03 100
92 0.14 0.08 0.10 100
93 0.08 0.07 0.07 100
94 0.63 0.32 0.42 100
95 0.21 0.30 0.25 100
96 0.21 0.24 0.22 100
97 0.07 0.07 0.07 100
98 0.06 0.16 0.09 100
99 0.48 0.33 0.39 100
accuracy 0.16 10000
macro avg 0.19 0.16 0.16 10000
weighted avg 0.19 0.16 0.16 10000
***************=================================================================***************